Service Assurance for the Virtualizing and Software-Defined Networks · Service Assurance for the Virtualizing and Software-Defined Networks ... Existing systems detect

Product Management, Cloud and Virtualization

October 7, 2015

Service Assurance for the Virtualizing and Software-Defined Networks Cisco Knowledge Network Presentation

Paola Arosio Deepak Bhargava

2 © 2014 Cisco and/or its affiliates. All rights reserved.

Agenda

NFV Adoption Trends

Required Attributes

Current OSS Limitations and Challenges

The New Approach to Service Assurance

Summary, Contact and Resources


NFV Adoption Trends

Source: Heavy Reading “NFV and Service Assurance Report”, June 2015


Current OSS Limitations and Challenges


OSS Limitations and Challenges for Service Assurance

§  Yet only 2% of users with a bad experience complain

§  73% of problems are reported by end users

73%

27%

98%

2%

By inference: §  Existing systems detect <1% of performance issues §  ~97% of issues are neither detected nor reported

Customer Reported

Issues

Provider detected issues

Customers who don’t complain

Customers who complain

Source: Forrester Study Source: Gartner “how to approach customer experience management”


NFV and SDN Introduce Further Complexities

Cloud, Elastic Compute, SDN and NFV Virtualized Networks and Compute

Transmission and Mobile

Application

Application NFV

NFV

Service

Service

“This stuff is complex. I don’t know if the fault is

me or them”

“We need to detect issues before the end

users call” “Everyone in Support is working in isolation of

each other”

Disassociated Apps & Infrastructure

DevOps

NfV

Apps

Disassociated Monitoring and Aggregation Tools

Alert Management / Log Management / APM / Performance Analytics

Disassociated People Silos

Owner Operated

Outsourcer Operated 3P Service


NFV and SDN Requires a New Approach to Service Assurance

CSP Business Requirements

Traditional Assurance Approach

Cisco’s Approach to Service Assurance (Aided by orchestration)

End-to-end service and customer experience focus

•  Resource centric •  Service state calculated as

an after-thought based on low-level KPIs

•  Service model based SLA decomposition and cross-domain data aggregation

Rapid Service Creation •  An after-thought •  Coupling of orchestration and assurance •  Service model-driven assurance

Visibility into dynamic Infrastructure

•  Bottoms-up modeling •  Rules-based approach

•  Subscription to VI/VNF changes and dependency mapping •  Self-healing and optimization enabled by feedback loop

Managing both virtual and physical resources

•  Static model for dedicated resources

•  Hybrid analytics & model based approach for service impact and root cause analysis

Seamless integration with OSS/BSS current environment

•  Segmented & specialized tools and operations

•  Modular and layered architecture delivering deployment flexibility •  Open horizontally scalable data platform •  Decoupling of publishing from consumption layer - break silo’s without

disrupting operations


Poll Question- What are your top three requirements for next-generation service assurance? (choose top 3)


A New Approach to Service Assurance


•  Self Healing: Policy-based automation that combine visibility and analytics to control and optimize

•  Out-of-Box Content: Pre-defined content for supporting use-cases

•  External Integration: Based on open APIs

Service Assurance: Key Tenets

x86

L2/L3 CPE (ISR, NID)

Hosted DC

vWSA vFW

vWAAS

Branch A

Branch B

VPN (IPSec or IPv6

L2TPv3

CPE Storage

Compute WAN NFV

CSR1kV NAT

•  Cross-domain and Multi-vendor: End-to-end visibility across multiple domains (i.e. CPE to WAN to NFV and Cloud)

•  Multi-layer: Correlated views across various layers - service, virtual, and physical

•  Orchestration Integration: Provision service assurance at the time of service instantiation

Deliver reliable services and a consistent user experience

Example: Cloud VPN


Modular Architecture Breaks siloes and enable vertical specialization

Service Management Problem/Incident Management

Service Health Impact Analysis

Service Quality Assessment

Optimization

Service/Workload Placement

Events Logs Metrics

Network, Compute, Storage (Physical, Virtual) Operations Business

Domain Specific

Cross-Domain

Capacity Planning & Forecasting

Fault and Cause Analysis

Customer Portal Operator Portal Executive Dashboard

Col

lect

ion

Ana

lysi

s P

rese

ntat

ion

Service Health Dashboard SLA Dashboard Routing &

Reservation Console

Event Analysis

Distribution Bus

Metric Analysis

BS

S/O

SS

M

ediation

Billing Mediation

Legacy OSS GWs

Loosely coupled and tight integration with Service Orchestration

Analysis layer covering specific capabilities: •  RCA, SIA, SLA •  Modular

approach to allow analytics plug-ins

Modular and Layered architecture allows for: •  Reuse of existing

collection mechanisms

•  Integration with existing customer and third-party OSS applications

Horizontal Scalable Platform to collect data from different sources and publish data to consumers

Out-of-box content to support specific use cases.

Use common YANG service model for service and assurance descriptors definition

Log Analysis

Cro

ss-D

omai

n O

rche

stra

tor

Fulfi

llmen

t

VMS VPC GiLAN Other …


On a scale from 1 (not important) to 5 (critically important), please rank the importance of having the following service assurance analytics functions in place when FIRST deploying NFV.

Poll Question: Which Service Assurance functions are gating your NFV deployment?


1.  Open and modular, aligned with Big Data framework

2.  Service model-driven assurance

3.  Analytics based OSS functions applied across physical and virtual infrastructure

Service Assurance Architecture Required Attributes


Open and Modular Architecture


Open and Modular Architecture Enables flexible deployment

Service Management Problem/Incident Management

Service Health Impact Analysis

Service Quality Assessment

Optimization

Service/Workload Placement

Events Logs Metrics

Network, Compute, Storage (Physical, Virtual) Operations Business

Domain Specific

Cross-Domain

Capacity Planning & Forecasting

Fault and Cause Analysis

Customer Portal Operator Portal Executive Dashboard

Col

lect

ion

Ana

lysi

s P

rese

ntat

ion

Service Health Dashboard SLA Dashboard Routing &

Reservation Console

Event Analysis

Distribution Bus

Metric Analysis

BS

S/O

SS

M

ediation

Billing Mediation

Legacy OSS GWs

Log Analysis

INSTRUMENTATION

NETWORK SERVICE MANAGEMENT

SERVICE ASSURANCE

Brownfield

Mix Brown-Green Field

Greenfield


§  Tight coupling of data aggregation/store/ analysis – pipelines realised in products

§  Multi-stage processing- both at aggregation and analysis

§  Architecture is a function of product choices

§  Filtering on aggregation to reduce data volume

§  Analysis functions largely based on programmatic rules, derived topology models and static thresholds

How Service Assurance is Realized Today

SNMP stats

Logs

SNMP traps

Data sources

Events

Data store

Perf Analysis

Fault Analysis

Data analysis

Stats

Log Index/Search Logs

Polling

Event aggregation

Data aggregation

Log Aggregation

Dashboard & Reporting


Outputs



OSS functions can be expressed as operations against the entire OSS data set: §  Fault management = ƒ(event data, metric data)

§  Performance management = ƒ(metric data)

§  Billing mediation = ƒ(event data, metric data)

§  Capacity management = ƒ(metric data)

Analytics in the Loop(s): Self-healing and optimization enabled by feedback loop

Big Data based Service Assurance Analytics enables self-healing and optimization


Big Data Architecture

Consumers: Data analysis Applications

Performance Analy.cs

Fault Analysis

SLA Repor.ng

Incident & Problem Management

Log Search

Capacity Analy.cs

Billing (Media.on)

Business Intelligence

Security and Threat Analysis

Orchestration

Controllers

Customer

Devices

Applications

QoE Monitoring

Infrastructure and

service-‐level data

Custom

er-‐le

vel

data

Data Distribution

Data Store & Processing

Master Data Store

Open Data Platform

Batch Processing

Stream processing

Live stream

Real Time Data Store

Dee

p H

isto

rical

Que

ry

Rea

l Tim

e Q

uery

Publishers: Data aggregation

Event aggregation

Log Aggregation

Metric aggregation

Network Telemetry

Benefits §  An open system architecture with no

dependency to any specific vendor or product

§  Allows any analytics application to mine any data source, leveraging the full value of the OSS dataset

§  Extensible – add new OSS analytics functions quickly and seamlessly with minimum development cost

§  Minimizes duplicate polling – collect data once, use many times

§  Remove cross-system integration

§  Leverage rapid innovation in Big Data analytics space

§  Platform is extensible beyond OSS


Service Assurance Functional Baseline

Metric Data

Event data

Real-time inventory

Performance Analytics

SLA Reporting

Incident & Problem

Management

Service Health

Real-Time Service Health

Time-series analytics

Event analytics

Views

Incident UI

Service Status Dashboard

Event Console

Orchestration

Controllers

Customer

Devices

Applications

QoE Monitoring

Infrastructure and

service-‐level data

Custom

er-‐le

vel

data

Data P

latform

Fault Analytics


Service Model Driven Assurance


Service Level Definition •  Service availability •  Loss, latency, jitter, …

Service Assurance •  Verify the service is available and

how it is performing •  Scale-up/-down based upon load •  Local recovery actions if the VNF

is unavailable/underperforming •  Identify underlying causes and fix

them asap Service Management &

Operations

Orchestration •  Put it there Service Provisioning

Service Assurance is a Service Lifecycle Problem

Service Placement

SLA Definition •  What SLA is required?

Service Monitoring Service Elasticity and Availability

Service Placement •  Where can it be supported?

•  Service Availability •  Monitoring •  Reporting

•  Admission Control •  Workload Placement

•  Service Elasticity and Availability

•  Performance Mgmt •  Service Level Monitoring •  Fault management {cause

analysis, Impact analysis} •  Incident / problem mgmt. •  Remediation

Day-0: Before service provisioning

Day-1: During service provisioning

Day-2: After service provisioning


SLA Definition Services are different…

§  No generic KPI measurement component for all service types

§  Smart components (i.e probes) per service-type often needed

§  Focus on Service-Type KPI definition and direct Service KPI monitoring

SLA Status •  Violated •  Jeopardized

Generic KPIs •  Latency •  Availability, Uptime •  …

Service-type specific KPIs




f()

Service Level Definition

e.g. MOS, App Response Time, MDI, vMOS

VoIP

Video on Demand

Video Streaming

Exchange

Sharepoint

WiFI

Cloud VPN

iWAN

IaaS


End to End Service Assurance

WAN

WAN

Managed CPEs Virtualized Data Center Infrastructure

Virtual Network FuncEons

WAN Service

IaaS #1

CPE Service VNF Service #1

End to End Service

IaaS #3

VNF #3

e2e Service

VNF Service IAAS Service WAN Service CPE Service

IaaS #2

VNF Service #2 VNF #4

SLA attributes (examples): Availability: 99.99% Response Time: < 100ms Throughput: > 1 Gbps

Network/Compute/Storage

Dashboard/Reports

Service, SLA & Policy Definition

Monitoring Policy Definition

Analytics (………..)

Cross-Domain, Multi-Vendor

Collection

Visibility-Policy-Control Feedback

Loop

Control/ Configuration changes

Orchestration Assurance

Domain Specific SLA

Decomposition (RFS)

Instrumentation Provisioning

Service Model Creation / Change

Feedback Loop

Service Provisioning


Service Assurance Auto Enablement Leverage YANG and Orchestration Engine for Service Assurance Provisioning

§  Extend service model to describe service intent and SLA descriptors

§  Leverage full power of YANG model to support the service assurance parameters

Dashboard/Reports

Service, SLA & Policy Definition

Monitoring Policy Definition

Analytics (………..)

Cross-Domain, Multi-Vendor

Collection

Instrumentation Provisioning

Domain Specific SLA

Decomposition (RFS)

Network/Compute/Storage

Service Model Extension

Configure collection systems

Auto-provisioning of instrumentation,

test capabilities and probes

Inventory Models to facilitate SIA

and RCA

Remediation feedback


Visibility-Policy-Control Feedback

Loop


Service Assurance Orchestration

NSO

VNFD

(V)NF VNFM Monitoring Systems Reporting

Defines scale and availability parameters for VNF which are managed locally by VNFM; determines VNF monitoring by VNFM

Defines VNF and associated SLA

Configures required instrumentation on that VNF, e.g. syslog, SNMP, etc.

Real-time inventory

Container {……. Leaf {…….

List {… } } }

Service Model


Configures reporting system so that it knows how to interpret monitoring data, e.g. roll-up calculations, alter thresholds etc.

Configures required monitoring on monitoring system for that VNF, e.g. what to poll for and when

Share Service Definition and infrastructure context for analytics

Activation Test: validates that the service works


Analytics based OSS Functions


Hybrid Analytics and Model-based approach Expedite Service Impact and Root Cause Analysis

WAN Service

IaaS #1

CPE Service VNF Service #1

IaaS #3

VNF #3

e2e Service

VNF Service IAAS Service WAN Service CPE Service

IaaS #2

VNF Service #2 VNF #4

Resource/Domain Manager



Model Based

Events Logs Metrics

Distribution Bus C

olle

ctio

n A

naly

sis

Cro

ss-D

omai

n O

rche

stra

tor

Service Model

Analytics Based

Dat

a E

nric

hmen

t


Events Filtered Alerts Situations Notifications Situation Rooms Network, Compute,

Apps, Sentiment

Real-time, Automatic Data Categorization and Noise Filtering

(Unsupervised Machine Learning)

Real-time, Automatic Anomaly Detection;

Alert Groups (Unsupervised Machine Learning)

Real-time, Automatic Situation Awareness;

Dynamic Teaming (Unsupervised Machine Learning)

Clean Contextualize Close

Analytics-based Fault and Situation Analysis Contextualized Alert Clusters reduces Time to Detect, Diagnose and Resolve


Analytics-based Fault and Situation Analysis Contextualized Alert Clusters reduces Time to Detect, Diagnose and Resolve

Events

Event Enrichment

Significance Ranking

Event Filtering Situations

Situation Enrichment

Situation Prioritization

Orchestration

Collaboration

Automation

Knowledge

Syslog Translation

SNMP Translation

Stakeholder Notification

NSO Ticketing System

Clean Contextualize Close


Analytics-based Fault and Situation Analysis Enhance Operational Efficiency with Collaboration and Knowledge base

Knowledge Base

Historical Algorithms Generated Similarity Factor Identifies Past

Similar Problems

User Comments Entered during troubleshooting and resolution of past

problems helps resolve current instance of recurring problem


Service Assurance: Attributes to effective operationalization of virtual infrastructures

An ideal solution must:

§  Leverage on YANG service models as a bridge between orchestration and assurance

§  Provide a horizontal and scalable big data platform & real-time data collection based on "publish/ subscribe" principles

§  Incorporate analytics functions across both physical and virtual elements

§  Enable self-healing through close looped feedback

Summary


Poll Question: Service Assurance in Hybrid Physical and Virtual Infrastructures


Contacts and Resources


Americas •  Moti Beharav: [email protected]

EMEAR •  Brett Holmes: [email protected] APJC •  Andrew Eaton: [email protected]

Resources

For More Information

Cisco Contacts Heavy Reading White Paper The Role of Service Assurance in the Virtualizing Network

Cisco Evolved Services Platform www.cisco.com/go/esp

Cisco Service Management and Orchestration Software Portfolio www.cisco.com/go/servicemano

l

Service Assurance for the Virtualizing and Software-Defined Networks · Service Assurance for the Virtualizing and Software-Defined Networks ... Existing systems detect

Documents